Source: intelligent things

Author: ZeR0

The aging traditional search engine is at the center of a new AI technology storm.

Just over the weekend, Microsoft Corp Bing, the world's second largest search engine, quietly launched a new version of Bing integrated with ChatGPT, and some lucky users have already tasted it.

Unlike traditional search engines, Bing's interface is not a slender search bar, but a larger chat box. After you enter your own question or what you want to inquire about, it will reply to you directly in the form of chat. At the same time, traditional search bar options are still available.

The user may be seeing an earlier version of the error. This shows the prototype of the functions of new AI chatbots like ChatGPT when combined with search engines. It is rumored that, unlike ChatGPT, which can only answer data before 2021, the Bing version will be able to access current information, and Microsoft Corp will officially release a new version of the improved Bing search engine in the next few weeks.

In addition to Microsoft Corp, search giants such as Alphabet Inc-CL C and Baidu, Inc. are also investing in research and development of ChatGPT competitors while preparing to launch a similar search engine "new species". According to rumors, Baidu, Inc. 's new search engine may be launched in March this year. Alphabet Inc-CL C will hold an AI event at 21:30 Beijing time on February 8th and may respond to ChatGPT's declaration of war.

In order to deal with the threat of ChatGPT, the two Alphabet Inc-CL C co-founders who have withdrawn from Alphabet Inc-CL C's daily business urgently returned to the company and initiated meetings many times to discuss countermeasures. Alphabet Inc-CL C has also invested about $300m in Anthropic, the AI startup that develops ChatGPT competitions. The founding members of Anthropic worked for OpenAI, which created ChatGPT.

In addition to threatening traditional search engines, ChatGPT is also shaking the cake for non-top programmers. According to an internal document obtained by foreign media, ChatGPT passed an interview with Alphabet Inc-CL C, a third-level engineer, with an average annual salary of $183000.

And engineers are trying to change the smart home experience with ChatGPT. Last month, Mate Marschalko, a senior Web developer, made a smarter voice assistant with Siri Shortcuts by interacting with the GPT-3 model behind ChatGPT in less than an hour. This voice assistant can not only control the entire Apple Inc HomeKit smart home system, but also easily answer all kinds of questions with ultra-low delay response.

He spoke highly of ChatGPT, saying that after trying the product, all the "smart" assistants, including Apple Inc Siri, Amazon.Com Inc Alexa and Alphabet Inc-CL C Home, looked so stupid and useless.

Unlike the "self-high" landing of AI technology in the past, ChatGPT is changing the shape of all mainstream Q & A products.

And this is just the beginning.

Microsoft Corp took it for commercial use to make money, and AWS asked it to write training documents.

If you haven't tried ChatGPT yet, it may be hard to understand why the AI product is so addictive to users around the world that you can't stop using it for two months.

To put it simply, because it is really very practical and can help ordinary peopleSave a lot of brainpower and time costs。

In terms of functionality, it can be a novel toy that arouses freshness, a chat master who spends boring time, an efficiency tool for productivity, and can be used as a treasure trove of knowledge about geography.

In just two months, people have been continuously tapping more skills of ChatGPT, including writing code, assignments, papers, speeches, event planning, advertising copywriting, movie scripts and other texts, or giving advice on home improvement design, programming debugging, life planning, and so on.

As long as you train carefully, ChatGPT can even rapidly evolve from a master of tricks to an efficient learning tool. After a series of questioning, it can list a large number of book lists and links to learning aids, help you extract the key points of an article, quickly grasp the core context of a field, or help you open up creative ideas when your inspiration dries up.

When a "new species" that subverts productivity and even causes a lot of professional crisis appears, all the tech giants are eyeing covetously.

The one who took the biggest step was Microsoft Corp. Although ChatGPT is not perfect, occasionally certain to answer some plausible rumble, or seems to be true but wrong content, but Microsoft Corp CEO Nadella with a big stroke of the pen: integrate ChatGPT in all products.

This high-profile declaration has been a great success, making Microsoft Corp, who has been silent in the field of cutting-edge AI technology for a long time, once again become the focus of the global science and technology circle. Its radical layout will answer a closely watched question-which Internet products will ChatGPT change the logic of operation?

At present, Microsoft Corp has plans to integrate OpenAI technology and ChatGPT in a variety of products, including Azure cloud services, Office office family buckets, Teams collaborative conference software, Bing search engine, Design design software, Dynamics business software, etc.

In other words, Microsoft Corp users will soon be able to get AI to write emails, manuscripts, meeting notes and other complicated and repetitive standard text work. It is also reported that Microsoft Corp may access a large number of AI applications in the Windows 12 operating system, which will be launched in 2024.

Microsoft Corp has already used Azure OpenAI services to power his automatic programming tool GitHub Copilot. ChatGPT makes it easier than ever to automatically program and check bug. As long as you write your own ideas in English, AI will be able to send the complete code to you. Even Andrej Karpathy, the former head of Tesla, Inc. AI, lamented on Twitter: "English is now the hottest new programming language."

Other large technology companies also get a piece of the pie.

According to foreign media reports, AWS, the Amazon cloud computing department, set up a small working group and found that ChatGPT did a "very good" job in answering AWS customer support questions, job interview questions, company strategy questions and writing training documents.

Meta, an American meta-universe giant, will use the American digital media companyBuzzFeedBenefit from OpenAI's technology. With the help of OpenAI API,BuzzFeed and AI to help create personalized content, the stock price soared. Meta recently paid BuzzFeed millions to help its social software generate content.

According to China Business report, he Xiaodong, vice president of JD.com Group and IEEE Fellow, said that JD.com will continue to integrate ChatGPT methods and technical points into products and services to promote the landing of AI industry.

No matter how the tech giants enter the market, once ChatGPT wins a commercial victory, Microsoft Corp will most likely be the biggest winner among these large technology companies.

Microsoft Corp's loss and ambition behind a huge bet on OpenAI

Last month, Microsoft Corp announced a huge additional investment of billions of dollars in OpenAI. According to US media Fortune, Microsoft Corp had already invested in OpenAI.3 billion dollars。

Microsoft Corp attaches so much importance to OpenAI, does not hesitate to continue to inject huge capital in the face of macroeconomic headwinds, and is so eager to apply the advanced technology of OpenAI to Microsoft Corp's products.

The answer may have something to do with Microsoft Corp's failure in commercializing AI research over the years.

Microsoft Corp Research Institute has been in the forefront of AI research for a long time, but for more than a decade, whether it is releasing breakthrough AI achievements that shocked the industry, or realizing the commercial transformation of AI research, Microsoft Corp seems to be half a beat behind other large technology enterprises.

When the brothers DeepMind and Alphabet Inc-CL C opened the landing chapter of deep learning and autopilot, when Amazon.Com Inc's intelligent hardware ecology based on Alexa voice assistant grew like a vine, when Apple Inc Siri voice assistant, text recognition and personalized recommendation technology made the operation of smartphones more convenient than ever, when Alphabet Inc-CL C BERT launched a global hurricane of large-scale language model research. Microsoft Corp is no longer the first name that comes to mind when people list AI's business achievements.

But Nadella is clearly not reconciled.

In fact, Microsoft Corp started early research on privacy related to AI as early as 2006. Since 2010, the department responsible for the research and development of speech processing and language understanding has been led by Peter Lee, vice president of Microsoft Corp Research Institute. These functions have been used in Office office software and Bing search engine to provide technical support for dialogue transcription, grammar checking and other functions. Since then, Microsoft Corp has been using advanced AI technology to imperceptibly improve the user experience of all kinds of products.

However, Microsoft Corp Research Institute, which focuses on the next revolutionary technological research, focuses more on research and development itself and turns a deaf ear to commercial value, so its research breakthroughs are not easy to be commercialized directly.

While Nadella is worried about Microsoft Corp's narrow defeat in AI, an once-in-a-lifetime opportunity lies ahead of him-OpenAI, which has developed many complex AI tools in just a few years, is looking for financing opportunities.

Under the command of Nadella, Microsoft Corp invested $1 billion in OpenAI in 2019 and bought out the basic technology behind GPT-3 in 2020.Exclusive licenseSince then, it has established a growing relationship with OpenAI and received quite a lot of preferential treatment, including priority licensing of most of OpenAI's technologies for products such as Bing search engines, and Azure cloud platform to become the exclusive cloud provider of OpenAI.

The new alliance soon bore fruit. Not long after OpenAI released the GPT-3 model, Microsoft Corp quietly integrated GPT-3 into his own products, such as improving Office's automatic word completion and optimizing voice-to-text tools. A senior sales manager of Microsoft Corp said that Microsoft Corp launched the OpenAI function much faster than the tools that worked on AI development at Microsoft Corp Research Institute before.

OpenAI is more and more inseparable from Microsoft Corp's software and hardware. According to foreign media sources, the OpenAI model takes up more space and computing in Azure than the Microsoft Corp language model Turing-NLG.

For compute-intensive development such as large language models, cost reduction is key. Assuming that there are 10 million users per month, the cost of running ChatGPT per day is estimated to be as high as $1 million. Nadella has revealed that Microsoft Corp has built a supercomputer to handle OpenAI work and can handle some AI calculations at half the cost of competitors.

In addition, since Microsoft Corp is the largest investor in OpenAI, the more commercial compensation OpenAI receives from ChatGPT after OpenAI launches its $20-a-month ChatGPT Plus subscription service, it means that Microsoft Corp can get a greater return. OpenAI expects revenue to reach $200 million this year and next year.1 billion dollars。

According to foreign media reports, Microsoft Corp can hold more than 3 shares of OpenAI at most. Once OpenAI starts to make a profit, after returning First close partners (FCP), 75% of the profit will be returned to Microsoft Corp and 25% to employees and FCP. Then, 2 per cent to OpenAI non-profits, 41 per cent to employees and 8 per cent to FCP,49% to Microsoft Corp. When all parties reach the income ceiling, 100% of the equity will be returned to OpenAI's non-profit.

However, the financial return is only a small profit, and the really great potential value lies in the large number of ChatGPT users gathered in a short period of time, and the trial feedback from millions of users snowballed OpenAI's first-mover advantage, prompting it to become the rulemaker of ChatGPT products.

It is not difficult to understand why Bill Gates, founder of Microsoft Corp, who has "turned a deaf ear to the company and focused on philanthropy" for a long time, also spoke highly of AI technology like ChatGPT, which is as important as PC and the Internet.

Today, with the rapid breakthrough of OpenAI research and development, Microsoft Corp's ambitions can no longer be compared with each other.

Microsoft Corp wants to add more advanced functions based on GPT-3.5 and GPT-4, including ChatGPT, to Azure, Office, Teams, Bing and other products, so as to continue to dominate productivity tools in the information age.

In order to achieve these goals as soon as possible, Microsoft Corp has transferred OpenAI researchers to Microsoft Corp's product group to speed up commercial revenue generation.

Alphabet Inc-CL C: invest in OpenAI competitors, test ChatGPT competitions, and design new search pages

OpenAI behind ChatGPT has aroused the vigilance of large science and technology enterprises all over the world because of its amazing creativity and talent density.

The company, which developed landmark AI products such as GPT-3 and ChatGPT, had a mere 375 employees as of January 24. Only two months after its launch, the number of monthly active users of ChatGPT has beenBreak through the 100 million markIs the fastest consumer application in history

The first to bear the brunt is Alphabet Inc-CL C, who meets Microsoft Corp in the fields of cloud computing, AI and search engines. To that end, Alphabet Inc-CL C founders Larry Page and Sergey Brin, who have left the company since 2019, returned to the company last month for a series of executive meetings to emphasize the urgency of dealing with ChatGPT.

Interestingly, Alphabet Inc-CL C seems to be going to imitate Microsoft Corp's strategy.

Alphabet Inc-CL C has invested about $300m in Anthropic, the AI start-up that develops ChatGPT competitions. Anthropic was founded in 2021, and its founder, Dario Amodei, was the vice president of research and development for OpenAI and left because of disagreements over the direction of OpenAI.

At present, the size of the Anthropic team is about 40, and most of the members have been involved in the research and development of GPT-2 and GPT-3 models, with the same goal as OpenAI-to build a reliable, explainable and controllable general artificial intelligence system.

Earlier this year, Anthropic unveiled its chat robot Claude, which directly targets ChatGPT. Alphabet Inc-CL C is currently its only publicly disclosed partner.

Before Alphabet Inc-CL C joined the investment, Anthropic had raised more than $700m, including $124 million in round A financing and $580 million in round B financing. According to foreign media reports, the latest $300m investment is led by Alphabet Inc-CL C Cloud and signed a large cloud contract with Anthropic to deploy Claude on Alphabet Inc-CL C Cloud. When the deal is completed, Alphabet Inc-CL C will hold about 10 per cent of Anthropic.

Similar to Microsoft Corp's OpenAI Alliance, this partnership is a win-win. Cloud computing giants can learn more about the technology and talent of startups, and startups don't have to spend a lot of money on building infrastructure. Amazon.Com Inc also formed an alliance with AI painting Unicorn Stability AI, and AWS became the "preferred cloud partner" for the startup to build and train AI content generation models.

The difference is that Microsoft Corp and OpenAI signed an "exclusive supply", the relationship is more closely bound, while Alphabet Inc-CL C and Anthropic, Amazon.Com Inc and Stability AI transactions are not exclusive, the two start-ups are free to cooperate with other cloud providers. However, considering that building a large model requires stable hardware, AI startups usually do not want to migrate between different cloud platforms after choosing a cloud platform.

Alphabet Inc-CL C recently tested a number of AI chat robots, including ChatGPT, according to an internal document. According to the answers, Alphabet Inc-CL C's team believes that the level of ChatGPT has reached the level of level III engineering positions, with an average total salary of about $183000.

Google Cloud is launching a "Red Alert" project for Atlas to deal with the threat of ChatGPT. Another product department has been testing a new search page that can be used in question-and-answer form. In addition, Alphabet Inc-CL C is also testing a chat robot Apprentice Bard that uses Alphabet Inc-CL C's dialogue AI language model LaMDA.

Apprentice Bard and ChatGPT function is similar, after the user enters the question in the dialog box, can give the detailed answer to the corresponding question in the form of more like human question and answer. And like the new version of Bing embedded in ChatGPT, Apprentice Bard is said to be able to answer recent events.

However, the reliability of its answer still needs to be improved. In an example circulating within Alphabet Inc-CL C, a tester asked Apprentice Bard: will Alphabet Inc-CL C make another round of layoffs?

It replied: "according to my database access, Alphabet Inc-CL C is unlikely to make another round of layoffs in 2023." Layoffs are usually aimed at reducing costs and structure, but the company is in good financial shape. In fact, Alphabet Inc-CL C's income increased by 34% in 2021, and the company's share price has risen by 70% since January 2022. "

The truth, however, is that in January 2023, Alphabet Inc-CL C announced 12000 layoffs, accounting for 6 per cent of his workforce.

Alphabet Inc-CL C, who has been repeatedly involved in AI moral public opinion, is obviously more cautious about the release of new products. Jeff Dean, head of Alphabet Inc-CL C AI, has previously told employees that Alphabet Inc-CL C has the ability to make products comparable to ChatGPT and is reluctant to release them because he is worried that such products will affect the company's goodwill due to defects such as providing misinformation, so they are more conservative than "small start-ups".

The series of troubles involved by ChatGPT also confirms that these concerns are by no means unfounded. Stack Overflow was forced to announce that ChatGPT was disabled because incorrect answers generated by ChatGPT were transported in large numbers to programming question-and-answer communities such as Stack Overflow. In addition, many students try to use ChatGPT to cheat or write papers, which is a headache for teachers in universities, primary and secondary schools in the United States.

As Werner Vogels, chief technology officer of AWS, puts it: "it (ChatGPT) doesn't care about the truth, it just combines convincing words."

But the urgent situation has forced Alphabet Inc-CL C to wait no longer. "We will carry out this work boldly, but with a strong sense of responsibility." Sandal Pichai, CEO of Alphabet Inc-CL C's parent company Alphabet, said Alphabet Inc-CL C will launch a large ChatGPT-like language model, LaMDA, "in the coming weeks or months," and users will soon be able to use the model in the form of "searching for partners."

After all, as the world's largest search engine giant and a pioneer in the AI field, Alphabet Inc-CL C may find it hard to watch competitors shake their own interest cake with an innovative AI product, and it's hard not to mind that the voice of an advanced AI technology is in the hands of competitors.

Conclusion: welcome to the era of generative AI

If you go back to a year ago, the AI industry is still frosty, and the R & D costs that continue to invest heavily but are difficult to generate put heavy pressure on technology giants and startups.

AI Healthcare pioneer IBM Watson Health has laid off a large number of employees, Tesla, Inc. and Waymo have made slow progress in fully autopilot, and the leading companies in computer vision and voice processing have not found a way to bring huge business returns.

All the changes began last year. Generative AI products, represented by AI painting and ChatGPT chat robot, are injecting a new dawn of commercialization into the AI industry by virtue of their strong strength to replace human painters and knowledge question and answer experts.

Painters, programmers and journalists are all wary of the popularity of this technology. This subversive AI work brings exciting productivity leaps, but it also brings some unsettling potential risks.

On the one hand, generative AI has the possibility of plagiarizing articles, code, images, videos and other human works; on the other hand, once it is used by unscrupulous people, it will become a "button" that will cause confusion and danger to the world, greatly reducing the threshold for making false information, cheating, falsifying images and even videos.

With the commercial use of generative AI, Pandora's AI box has been opened and everything is full of unknowns.

Edit / Viola